[DNM] [CS] Handle holes in mergeEquivalenceClasses
#83724
Conversation
|
@swift-ci please test |
|
@swift-ci please test source compatibility |
| let x = foo([], i) | ||
| // expected-error@-1 {{generic parameter 'T' could not be inferred}} | ||
| // expected-error@-2 {{cannot convert value of type '[Any]' to expected argument type '_.Type'}} | ||
| // FIXME: The above should say `T.Type` not `_.Type` ^ |
There was a problem hiding this comment.
I think this is a bug in FailureDiagnostic::resolveType, might be easy to fix.
There was a problem hiding this comment.
Yeah the issue is that we need the hole's locator, but that currently requires iterating over the equivalence class, which we don't have in the Solution. So either we need to stick the locator on the type variable itself as an extra bit of state, or stick a mapping in the Solution
There was a problem hiding this comment.
Should we just make that transformation part of ConstraintSystem::finalize then?
There was a problem hiding this comment.
The actual type being diagnosed is part of the fix though, right? Unless I'm misunderstanding what you mean by transformation
There was a problem hiding this comment.
We never want to print type variables so maybe replacing them with an underlying generic parameter should be done as part of forming a solution instead of delaying it to diagnostics like we do right now.
There was a problem hiding this comment.
Ah right I see what you're saying, let me give that a try
No description provided.